https://www.cell.com/cell-reports/fulltext/S2211-1247(26)00504-8
chatGPT(5.6paid)
Paper
Zhang et al. (2026), “Green light relieves stress-induced anxiety-like behaviors via visual-to-prefrontal projections in mice,” Cell Reports 45, 117426.
Executive summary
The paper reports that exposure to approximately 515 nm green LED light at 200 lux reduces anxiety-like behaviour in acutely stressed mice. The authors propose a specific neural circuit:
[
\text{Green light}\rightarrow \mathrm{V2M^{Glu}}\rightarrow \mathrm{PrL^{CRF}}\rightarrow \mathrm{PrL^{Glu}}\downarrow
]
Green light activates glutamatergic neurons in the medial secondary visual cortex (V2M). These project to corticotropin-releasing-factor-expressing inhibitory neurons in the prelimbic cortex (PrL^CRF), which suppress anxiety-associated glutamatergic pyramidal neurons (PrL^Glu). CRF receptor 1 (CRFR1) signalling contributes to this local inhibition.
The study combines behavioural assays, viral tracing, calcium imaging, slice electrophysiology, chemogenetics, optogenetics and local pharmacology. Collectively, these experiments provide reasonably strong evidence that this visual-to-prefrontal circuit is involved in the effect in the authors’ mouse stress paradigms.
The evidence is much weaker, however, for the broader proposition that green wavelength itself is uniquely anxiolytic, or that green-light treatment is ready for translation into an anxiety therapy.
What the authors did
Stress and behavioural models
The principal model was 30 minutes of acute restraint stress (ARS). A second model exposed mice to a 50°C hot plate for five minutes. The authors also conducted a seven-day chronic restraint experiment.
Anxiety-like behaviour was measured using:
- time and entries into the centre of an open field;
- time and entries into the open arms of an elevated-plus maze;
- a nominal conditioned-place-preference experiment.
Green light was tested at 50, 200 and 500 lux, with 200 lux producing the clearest behavioural effect. At nominally matched illuminance, white and blue light did not reproduce the effect.
Importantly, green light worked when delivered during stress induction but not when delivered only during the subsequent recovery period. This suggests that it interfered with the formation or encoding of the stress response, rather than accelerating recovery from an already established anxious state.
Prelimbic glutamatergic neurons
Acute stress increased calcium activity in PrL glutamatergic neurons. Green light reduced this hyperactivity, whereas blue and white light did not.
Chemogenetic inhibition of PrL glutamatergic neurons itself reduced anxiety-like behaviour. This supports the idea that suppression of these neurons is functionally relevant rather than merely correlated with green-light exposure.
Identification of the visual input
Retrograde and trans-synaptic viral tracing identified projections from the medial secondary visual cortex to the prelimbic cortex. Approximately 90% of retrogradely labelled V2M neurons reportedly expressed a glutamatergic marker.
Anterograde tracing suggested that V2M inputs contacted two main PrL populations:
- approximately 70% CRF-positive neurons;
- approximately 30% glutamatergic neurons.
Green light activated a larger proportion of V2M glutamatergic neurons than blue or white light in the authors’ imaging experiments.
The proposed feedforward circuit
Optogenetic activation of V2M glutamatergic neurons:
- rapidly activated PrL^CRF neurons;
- produced a net reduction in PrL^Glu calcium activity;
- generated direct excitatory postsynaptic currents and delayed inhibitory postsynaptic currents in PrL^Glu neurons.
The longer latency of the inhibitory currents—about 13 ms versus 5.6 ms for excitation—supports a disynaptic feedforward arrangement:
flowchart TD
A["Green light"] --> B["V2M glutamatergic neurons"]
B --> C["PrL CRF interneurons"]
C --> D["GABAergic and CRF-dependent inhibition"]
D --> E["Reduced PrL pyramidal activity"]
E --> F["Less anxiety-like avoidance"]
Activation of the V2M-recipient PrL^CRF population reproduced the neural and behavioural effects of green light. Inhibition of the same population worsened anxiety-like behaviour and increased PrL^Glu activity.
CRF/CRFR1 involvement
Optogenetic activation of PrL^CRF neurons evoked inhibitory currents in PrL glutamatergic neurons. The CRFR1 antagonist NBI27914 partially reduced these currents.
Local infusion of NBI27914 into the PrL also abolished the behavioural benefit of green light. The authors therefore argue that CRF release and CRFR1 activation contribute to the suppression of PrL pyramidal neurons.
This is not presented as purely peptidergic transmission: PrL^CRF cells are GABAergic interneurons, and the CRFR1 antagonist only partially suppressed the inhibitory currents. The most plausible interpretation is that conventional GABAergic transmission provides much of the immediate inhibition, with CRF/CRFR1 signalling modulating or strengthening the circuit.
Main conclusions
The results support four reasonably distinct conclusions:
-
Green-light exposure can reduce avoidance behaviour produced by acute restraint or noxious heat stress in mice.
-
Acute stress is associated with hyperactivity of PrL glutamatergic neurons, and green light reduces that hyperactivity.
-
V2M provides direct excitatory input to both PrL^CRF interneurons and PrL glutamatergic neurons, but recruitment of local inhibition makes the net effect on the latter inhibitory.
-
The V2M→PrL^CRF circuit and local CRFR1 signalling are necessary for the full behavioural effect under the tested conditions.
Novelty
1. Identification of a visual-cortical-to-prefrontal anxiolytic pathway
The strongest novelty is the identification of a previously poorly characterised V2M→PrL pathway through which visual input can regulate stress-related behaviour.
Previous work had established that light influences mood, pain and circadian functions, but much of that literature emphasised non-image-forming retinal pathways involving melanopsin, the hypothalamus, habenula or other subcortical structures. This paper instead places a higher-order visual cortical area at the centre of the effect.
2. Cell-type-specific feedforward inhibition
The paper goes beyond regional activation and proposes a defined cellular circuit:
- V2M glutamatergic neurons excite PrL CRF-expressing interneurons;
- those interneurons inhibit local glutamatergic pyramidal neurons;
- suppression of the pyramidal population reduces avoidance behaviour.
The combination of projection-specific tracing, calcium recording and causal manipulation makes this more novel and informative than merely showing activation of V2M or PrL.
3. CRF as a local resilience signal rather than simply a stress signal
CRF is usually discussed in relation to hypothalamic stress activation and the HPA axis. Here, CRF-expressing prefrontal interneurons appear to have an anti-anxiety or stress-resilience function.
That is biologically interesting, although it builds upon earlier studies showing that dorsal-mPFC CRF neurons can promote active coping and inhibit pyramidal cells.
4. A mechanism acting during stress induction
The distinction between treatment during restraint and treatment after restraint is important. The apparent effect on stress-response formation, rather than post-stress recovery, provides a more precise mechanistic and potentially clinical hypothesis.
5. Integration of spectral input with cortical emotional regulation
The authors link a nominally wavelength-selective visual response to a particular prefrontal microcircuit. That conceptual integration is novel, even though, as discussed below, the experiments do not yet establish that wavelength is the decisive variable.
Strengths
-
The paper uses several independent methodological levels: anatomy, population calcium signals, single-cell calcium imaging, electrophysiology, chemogenetics, optogenetics, pharmacology and behaviour.
-
It includes both necessity and sufficiency experiments. Activating the pathway mimics green light, while inhibiting it or blocking CRFR1 reduces the effect.
-
Locomotor distance was generally unchanged, reducing the likelihood that altered open-field and elevated-maze performance simply reflected sedation or motor activation.
-
The TTX/4-AP electrophysiological experiments support direct monosynaptic V2M input.
-
The different latency of excitatory and inhibitory currents provides useful physiological evidence for feedforward inhibition.
-
The absence of a clear effect in unstressed animals suggests some state dependency rather than indiscriminate suppression of normal behaviour.
-
The second acute-stress model and preliminary chronic-restraint experiment improve generalisability within rodents, although the circuit work remains centred on acute restraint.
Critique
1. The study does not rigorously isolate wavelength from retinal stimulation
This is the most important limitation.
The lights were matched at approximately 200 lux, but lux is weighted according to human photopic visual sensitivity, not mouse retinal sensitivity. It is therefore not a valid way to equate the effective biological intensity of different wavelengths in mice.
The reported physical exposures also differed:
| Light | Peak | Reported photon measure |
|---|---|---|
| Green | 515.1 nm | (1.08\times10^{12}) quanta/cm² |
| Blue | 455.5 nm | (5.64\times10^{11}) quanta/cm² |
| White | peak around 449.4 nm | (2.54\times10^{12}) quanta/cm² |
Green delivered approximately twice the reported photons of blue and lies close to the sensitivity maxima of mouse rhodopsin and M-opsin. Thus, green may simply have produced stronger effective activation of the dominant mouse photoreceptor system.
The white-light result argues against a simple total-photon explanation, but white light is spectrally broad and cannot be characterised adequately by one “peak wavelength.” Its effective excitation of rods, S-opsin, M-opsin and melanopsin should have been calculated separately.
A convincing wavelength-selectivity experiment would equate:
- retinal irradiance;
- photon flux within narrow spectral bands;
- rod, cone and melanopsin effective excitation;
- pupil size and retinal adaptation;
- perceived brightness for the mouse visual system.
Until that is done, the conclusion should be framed as “the tested green-light condition was most effective”, not “green wavelength is intrinsically anxiolytic.”
2. No retinal pathway to V2M is demonstrated
The study starts mechanistically at V2M. It does not establish:
- which photoreceptors detect the beneficial light;
- which retinal ganglion-cell subtype carries the signal;
- whether the signal reaches V2M through the classical retinogeniculate pathway, superior colliculus or another route;
- whether melanopsin contributes;
- whether colour-opponent retinal processing is required.
The proposed explanation involving rod/M-cone sensitivity and green-selective retinal circuits is therefore plausible but speculative.
Photoreceptor-specific knockout mice, retinal silencing, LGN/superior-colliculus manipulations or projection-specific tracing would be needed to complete the pathway.
3. The “conditioned place preference” experiment is not a strong test of anxiolysis
The experiment does not appear to be a conventional conditioned-place-preference design. The mice first explored with the light off and were then tested after the light in a fixed right-hand compartment was switched on.
Potential problems include:
- no clear counterbalancing of the illuminated side;
- order was apparently always no-light followed by light;
- preference could reflect visual attraction, phototaxis, novelty or altered exploration;
- there was no conditioning phase demonstrating learned association with relief;
- time-dependent recovery from stress could be confused with the second observation period.
It is therefore better described as an acute light-compartment preference test, and even then it does not independently demonstrate that the light was experienced as anxiolytic.
4. “Anxiety” is inferred from limited behavioural measures
Open-field-centre and elevated-plus-maze behaviour are sensitive to anxiety-like avoidance, but also to:
- threat perception;
- risk assessment;
- exploratory drive;
- visual acuity and contrast;
- arousal;
- stress-induced freezing;
- changes in sensory processing.
Because the intervention itself is visual and the assays depend on visual perception and exploration, additional behavioural validation is particularly important.
Useful additions would include the light-dark box, novelty-suppressed feeding, acoustic startle, physiological stress measures, corticosterone and autonomic readouts.
5. The acute-stress models have limited clinical correspondence
A 30-minute restraint model is not equivalent to an anxiety disorder. The hot-plate condition may additionally involve pain modulation: green light is already reported to have antinociceptive effects. Reduced pain could therefore indirectly change behaviour after hot-plate exposure.
The findings presently support effects on short-lived stress-induced avoidance, not established generalized anxiety, panic disorder or chronic pathological anxiety.
The seven-day restraint result is useful, but the circuit-level causal experiments were not replicated extensively in the chronic model.
6. Sex reporting is inconsistent and generalisability is poor
The main methods state that only male mice were used “for the entire study,” and the discussion refers to male mice. However, the main Results initially say that male and female mice underwent restraint, and Figure S1 reportedly contains an experiment in ARS female mice.
This needs clarification. If females were tested only in one preliminary behavioural experiment, that should be stated explicitly, along with estrous-cycle handling and sex-by-treatment analysis.
The explanation that males were chosen to avoid hormonal fluctuations is not sufficient justification for excluding females from a study of anxiety, particularly because anxiety disorders have substantial sex differences.
7. CRFR1’s mechanistic role is not completely resolved
The authors sometimes imply a direct chain in which CRF activates CRFR1 and thereby inhibits PrL glutamatergic neurons. But CRFR1 is usually a Gs-coupled receptor and can have context-dependent excitatory or modulatory effects.
The electrophysiological evidence shows that:
- PrL^CRF cells are inhibitory interneurons;
- activating them produces IPSCs in PrL^Glu neurons;
- a CRFR1 antagonist only partially reduces those IPSCs.
This suggests combined GABAergic and peptidergic mechanisms, not a simple CRF-mediated inhibitory synapse. The paper does not establish:
- whether CRFR1 is located directly on the recorded pyramidal neurons;
- whether it is present presynaptically on GABAergic terminals;
- whether CRF alters GABA release;
- whether CRFR2 contributes;
- the relative contribution of CRF versus GABA.
Cell-specific CRFR1 deletion would be considerably more decisive than local pharmacological blockade.
8. Some wording is internally contradictory
The discussion states that pharmacological blockade of CRFR1 “suppresses PrL^Glu neuronal activity,” but the reported electrophysiology shows that NBI27914 increased firing, meaning that CRFR1 blockade disinhibited these neurons.
This is likely an editorial error, but it occurs at the centre of the proposed mechanism and should have been corrected.
9. Pharmacological controls are incomplete
The authors acknowledge that CRFR1 blockade was not tested sufficiently in:
- unstressed animals;
- stressed animals without green light.
Consequently, abolishing the green-light behavioural effect does not necessarily mean that CRFR1 selectively transduces green-light anxiolysis. The antagonist may independently increase stress sensitivity or baseline avoidance.
A proper interaction experiment would compare stress, light and antagonist in a fully factorial design.
10. Chemogenetic and viral specificity remain concerns
CaMKII-based targeting is commonly used to enrich for excitatory cortical neurons, but it is not absolutely cell-type-specific. Likewise, trans-synaptic AAV methods can vary in transport efficiency and specificity.
CNO-based DREADD experiments can also be affected by conversion to clozapine. Virus-negative/CNO-treated controls appear to have been used in several experiments, but systematic vehicle, CNO and receptor-expression controls should be explicit for every behavioural manipulation.
11. Possible pseudoreplication is difficult to assess
The statistical section says that experiments included at least five mice with multiple brain slices. It is not always clear whether the statistical unit was:
- the mouse;
- the slice;
- the recorded cell;
- an individual calcium event.
Treating multiple cells or slices from the same animal as independent observations would artificially inflate statistical power. Hierarchical or mixed-effects modelling would be more appropriate for nested imaging and electrophysiological datasets.
Reporting mean ± SEM and thresholded significance symbols also makes effect size and uncertainty harder to judge. Exact (p)-values, confidence intervals, animal-level plots and prespecified primary outcomes would strengthen the work.
12. Data transparency is below current best practice
The paper says that data are available from the lead contact upon request and provides no deposited original code. Given the complex calcium-imaging, cell-registration and circuit analyses, public availability of:
- raw behavioural data;
- calcium traces;
- cell-classification thresholds;
- image-analysis scripts;
- electrophysiological recordings;
- viral-placement exclusions
would materially improve reproducibility.
Overall assessment
This is a technically ambitious and mechanistically interesting mouse circuit paper. Its strongest contribution is not the general observation that green environments or green light can be calming. Rather, it is the identification of a plausible V2M→PrL^CRF→PrL^Glu feedforward inhibitory circuit that can suppress acute stress-induced avoidance.
The circuit conclusion is supported by convergent evidence and is likely the paper’s genuinely important advance. The CRF/CRFR1 component is credible but incompletely localised and probably operates alongside conventional GABAergic inhibition.
The claim of a specifically green therapeutic mechanism is substantially less secure. Because the different lights were equated using human lux rather than mouse photoreceptor-effective irradiance, the experiment cannot cleanly distinguish colour from differential retinal drive. Nor does the study yet demonstrate the upstream retinal pathway or therapeutic efficacy in a chronic disease model.
My bottom-line interpretation would therefore be:
The paper convincingly identifies a visual-cortical/prefrontal circuit capable of limiting acute stress-related avoidance, and shows that the tested 515-nm light condition recruits this circuit particularly effectively. It does not yet establish that green wavelength per se is uniquely anxiolytic or that green-light exposure is a validated treatment for human anxiety disorders.